A Study Of Particle Swarm Optimization Algorithm Based On Migration Mechanism And Its Applications

Compared with the traditional optimization methods,the swarm intelligent optimization algorithm has no continuity requirements for the requested problem,has a strong adaptability to uncertain data,and just need through the simple evolution of information interaction between individuals can get the optimal solution to the problem.Therefore,it is widely used to handle complex optimization problems.Particle swarm optimization algorithm(PSO)is one of the important branches of the swarm intelligent optimization algorithms,which except has the advantages of swarm intelligent optimization algorithms,also has the advantages of simple tune parameters,strong search ability,and high robustness of algorithms.Therefore,it has received extensive attention,and applied in various fields.However,as with other intelligent optimization algorithms,the traditional PSO algorithms still have the disadvantages of slow convergence rate,low convergence accuracy,and easy to fall into local optimum.The migration mechanism is a local search mechanism originated from the biogeography-based optimization algorithm(BBO),and has a fast local convergence.In addition,the migration mechanism also includes two-tier elite replacement and migration mutation operations,which can maximize the retention of optimal information while increasing the diversity of the swarm.Therefore,this paper introduces the migration mechanism into the latest standard PSO algorithm(SPSO2011),and proposes an improved SPSO2011 algorithm to solve the problem of complex function optimization and further apply it to the selection of information genes in gene expression profiles.This kind of improved PSO algorithm not only can improve the local convergence ability of PSO algorithm,but also can make the algorithm jump out of the local minimum point and effectively improve the performance of PSO algorithm.The main work of this paper is as follows:(1)For the SPSO2011 algorithm to deal with complex function optimization problems,there is a slow convergence speed and easy to fall into the local optimum and other problems,proposed a PSO algorithm based on the migration mechanism(ISPSO2011).Based on the SPSO2011 algorithm,this algorithm introduces a migration mechanism that is insensitive to swarm diversity and has strong local search capability to effectively develop particle search space.When the algorithm falls into a local optimum,the migration mutation operation can make it have a certain ability to jump out of the local optimum.In addition,in order to share solution features more effectively and improve the search ability of the algorithm,a topology migration operation is adopted in the migration mechanism so that the information sharing between the particles is more sufficient.At the same time,the two-tiered elite mechanism is used in the algorithm to retain better particles to improve the convergence speed of the algorithm.Experiments conducted on multiple benchmark test function sets show that the algorithm achieves better results in convergence accuracy and theoretical minimum success rate than SPSO2011 algorithm and BBO algorithm.(2)A multiple gene selection method based on ISPSO2011 algorithm is proposed(IMRF-ISPSO2011).It can quickly eliminate a large number of noise genes and redundant genes,and at the same time obtain a subset of genes with higher classification accuracy.The method firstly uses the Relief algorithm to initially screen the original gene expression profile dataset,and remove a large number of noise genes.When the traditional FCBF algorithm deals with datasets with high correlation between genes and low correlation between genes and categories,it is easy to eliminate too many information genes.A corresponding improvement has been made in the culling strategy,which is used to perform a quick secondary screening of the primaries,and the ELM classifier is used to select the secondary gene pool that meets the classification accuracy.Finally,the secondary gene pool was finalized using the ISPSO2011 algorithm.By comparing the classification results of the gene subsets before and after the introduction of the ISPSO2011 algorithm,the validity of the ISPSO2011 algorithm in gene selection was verified.In addition,compared with other traditional PSO algorithm-based gene selection methods,this method has obtained a more powerful and more compact subset of genes.